Dell Storage - "good enough" in an "as-a-service" world?

Last year I wrote an article discussing the challenges of Dell’s Storage portfolio and the ongoing refresh of technologies that have a 20+ year heritage. As we move into a new world of “as a service” delivery, does it matter that Dell (and for that matter, many other companies) have inferior products if the service delivery method obfuscates the implementation details of the solution?

Background

As discussed in this article from last year, Dell Technologies has a wide range of products that fit into two categories – they are strategic or moving towards end-of-life. As a portfolio-holding company, Dell (and formerly EMC) has made substantial revenue from acquisitions that include PowerScale (Isilon), PowerStore (Clariion Data General) and PowerFlex (ScaleIO). At the time of acquisition (or just after), the Dell/EMC products were classed as market leaders and best-of-breed. However, over time, new solutions have come to the market that arguably offered better features than those in the Dell/EMC stable.

At the same time, there have been a few Dell/EMC misses. XtremIO was never successful enough to displace PowerMax/VMAX or even the Unity platform. DSSD was arguably too early but could have been a strong player, eventually. End-of-life products include solutions like Compellent and Equallogic.

Journey

Once Dell had completed the EMC acquisition, the diversity of storage choices available to customers required some degree of rationalisation. In this post from 2017, we highlighted the disparity of four mid-range storage platforms in the Dell portfolio, all with subtly different features and functionality.

By 2019, the choice to develop a mid-range successor was a three-horse race in which the heritage of the Unity/Clariion platform won out. With so many customers already using Unity/VNX, it was hard to see how the result could be anything different, albeit with some features from other platforms being candidates for integration into the new PowerStore.

Features

At launch, PowerStore appeared less than impressive, with some work to do in order to catch up to the competition. However, PowerStore did fix some of the issues of Unity, moving to NVMe and introducing a hybrid model with the ability to run local applications (AppsON). Dell has continued to iterate the platform new incremental features and functionality.

APEX

In May 2021, Dell introduced APEX, a new model for “as a service” consumption of Dell technology. The idea is not new. As we discussed in this 2018 blog, infrastructure vendors have been evolving their offerings for some time. In this follow-up, we discuss the challenges and benefits in more detail. HPE recently announced Alletra, an approach that obfuscates the specific hardware platform behind a services model. Both Primera and Nimble are now just Alletra storage services.

With APEX, Dell is looking to take the same approach. Hardware solutions are now simply services delivered through a web portal using basic service requirements, including capacity, performance, subscription term and physical location. The first deployments of APEX offer storage and virtualisation services.

Public Cloud

Why are traditional infrastructure vendors so determined to sell their wares as services? The answer, obviously, is the impact we’ve seen from the adoption of the public cloud. Since AWS launched SQS in 2004, the public cloud has evolved at incredible rates to a market size of $371 billion in 2020 and a projection of $832 billion by 2025. This rate, if accurate, is a CAGR of 17% and something on-premises infrastructure providers can only dream about.

Over the past 20 years, the public cloud has evolved from a niche solution to the strategic first choice for many IT organisations. The initial FUD about reliability and security have been largely discounted, with many start-ups never deploying on-premises infrastructure and existing only in the public cloud.

It’s hard to estimate the drain on revenue the public cloud has caused. However, we can see some evidence in the way the traditional storage hardware market has stagnated in recent years (although the tracking process does need to evolve). At the outset of the public cloud, most data was stored in block-based solutions or file servers. Now, we see massive growth in unstructured data and the use of high-performance file stores. Unstructured data growth aligns nicely with a cloud model where the majority of content is generally inactive, and data is stored/retrieved through an API call.

However, the public cloud isn’t all about cheap storage. Cloud offers a highly flexible consumption model where resources can be used for hours, minutes or even seconds with serverless computing. Businesses don’t need to buy expensive technology such as GPUs when they can be rented by the hour in the public cloud. Once one technology becomes obsolete, consumers can just move on to the latest and greatest solutions. Cloud vendors continue to push the “service density” by providing more application-based solutions running on common cloud infrastructure.

Private Cloud

This change in consumption represents a massive challenge for Dell and many other traditional infrastructure providers. How does Dell solve the dichotomy of on-demand resource consumption with the dedication of resources to individual customers represented by on-premises deployments?

The problem is almost impossible to solve because the key attributes of the public cloud – scale and shared access – means resources can be assigned between customers dynamically and at zero cost. With “on-premises cloud”, the infrastructure is dedicated to the customer, immediately removing the benefits of flexibility and introducing additional costs to achieve the requirements of infinite scale.

Starting Somewhere

Ultimately, vendors like Dell and HPE must start somewhere. To date, the APEX offerings look more like financial instruments than technical ones. Customers don’t own the on-premises equipment but do pay through an ongoing usage charge. Theoretically, Dell and similar vendors could remove equipment that isn’t in use, although that’s unlikely to happen.

The compromise in the on-premises “as-a-service” offerings is with the commitment timescale. Customers must commit to a minimum 12-month term, and naturally, vendors will be keen to offer some discounts over time. This process is analogous to public cloud reserved instances and represents a cost-saving for reducing the risk incurred by the hardware provider.

Obfuscation

Looking back at the premise of this article, the question is whether Dell storage products are “good enough” in an on-premises “as a service” delivery model.

In the public cloud, customers are offered no insight into the technology used to deliver their services. Vendors have created bespoke designs that optimise many parts of the infrastructure to reduce costs and improve operational efficiency.

In contrast, Dell and other vendors are simply using the same products they sell as a capital purchase and using them to deliver as-a-service offerings. This approach introduces several problems.

Many solutions aren’t designed for the level of granular deployment needed in on-premises “as-a-service”. For example, capacity upgrades and downgrades may require entire RAID groups of shelves of disks to be deployed.
Many solutions aren’t designed for upgrade-in-place, resulting in data migration project work for the customer (at their cost and risk).
Many solutions require manual configuration and management to achieve the best level of performance and aren’t “autonomic”. This issue causes problems scaling infrastructure support when managing thousands of devices.
Most storage solutions are still using technologies (such as Fibre Channel) that continue to need manual management and don’t integrate easily with dynamic storage services.

Remember that the design of enterprise storage has developed from a centralised model of highly consolidated appliances based on storage area networking. This is different from the public cloud model, where storage is just another service on top of the infrastructure.

Dell Strategy

As the “as a service” model gains ground, does it matter that Dell products aren’t leaders in storage innovation? To answer that question, we have to ask how the consumption model of storage changes in the “as a service” world. As we mentioned earlier, the new metrics for on-demand consumption are location, capacity, performance, and connection type (block/file/object). From the customer’s perspective, Dell has to offer storage in multiple locations, with capacity on-demand across multiple performance tiers and access methods.

From Dell’s perspective, the company needs to ensure that meeting these metrics can be achieved in a cost-effective and timely fashion. This means:

Just-in-time deployments of hardware, based on the customer growth curve. This requires meeting the requirements across multiple tiers and multiple locations with fine granularity (e.g. an individual disk/SSD).
Non-disruptive hardware upgrades, preferably in place or with minimal data migration.
Remote fault diagnostics and repair. Minimal visits onsite for hardware replacement, balancing risk versus cost management.
Hardware flexibility – the ability to use hardware for any protocol connection type.

These requirements will apply equally to other Dell solutions, including servers and networking. The implication is that future Dell storage and infrastructure solutions need to be more focused on operational efficiency and simplicity rather than a deep level of functionality. In essence, future Dell products need to be more cloud-like than enterprise in design.

Green Shoots

How well does Dell stack up to the operational challenge? From a hardware perspective, the storage solutions have some way to go. Neither PowerMax nor PowerStore could be defined as “set and forget” solutions. However, PowerScale offers a software-defined route that could be more flexible in larger on-premises environments.

PowerStore offers a future vision that introduces some compute functionality, which could be used to deliver more complex solutions, such as database-as-a-service or managed infrastructure services like DNS/DHCP/AD.

Dell also has a partnership with Liqid that provides configurable on-premises data centres. This technology could replace traditional storage with on-demand storage such as the Liqid LQS4500 (aka Honey Badger). The Liqid framework also offers much greater efficiency in the use of other more expensive compute resources such as GPUs.

The Architect’s View™

Dell may not have the best storage portfolio available, but the company does have a huge install base of customers that will be looking to optimise their infrastructure costs. APEX version 1.0 should be about introducing customers to the flexibility of the consumption model without the need to understand the underlying hardware. APEX 2.0 can then be more about optimising on-premises infrastructure, replacing components with more operationally efficient solutions.

Ultimately, Dell requires “as a service” customers to have the same mindset as those who adopted CI a decade ago. Converged Infrastructure didn’t bring the best-of-breed but instead integrated the components of a single vendor into a unified stack that should have been more efficient to operate. APEX needs to deliver the CI experience and more without the new challenges CI introduced.

As APEX evolves, we will be watching the operational aspects as keenly as the technology evolution because the successful vendors will be those operating the most operationally and financially efficient on-premises platforms.